Senior Software Engineer | T-Mobile (US) at Concentrix Catalyst (2025-07 – Present)
- Architected and deployed an LLM-powered ticket-triage system using Azure OpenAI, ServiceNow, and Splunk, cutting analysis time from 10-30 minutes to under 5 minutes and automatically reassigning incidents to the owning team.
- Built a three-stage LangChain pipeline for field extraction, summarization, and SPL generation over multimodal ticket content (text, images, and JSON); implemented a two-hop Splunk flow from discovery logs to GUID resolution and transaction logs to route failures to the owning team.
- Engineered a feedback-driven RAG loop that embeds corrected ticket summaries and retrieves semantically similar resolved incidents to improve subsequent triage prompts.
- Architected and deployed an alert-triage platform processing ~15K production alerts per month using Azure OpenAI, LangGraph, Azure AI Search, Cosmos DB, and RabbitMQ, reducing typical SOP discovery, execution, and RCA time for new operators from ~10 minutes to under 5 minutes.
- Designed an agentic execution harness combining deterministic SOP workflows with LLM fallback and MCP integrations across Splunk, Grafana, Microsoft Teams, and Confluence to execute conditional investigations and expand diagnostic time windows when evidence is incomplete.
- Engineered a RabbitMQ-based asynchronous monitoring pipeline that re-evaluates alert conditions after configurable delays, removing manual tracking and saving an estimated 1-2 minutes of repeat-query effort per monitored alert while escalating deteriorating failures for automated investigation.
- Built an agentic SOP-authoring workflow that converts Confluence runbooks into executable workflows, validates and repairs them against operational data, and incorporates human approval and replay-based regression checks before production use.
AVP Data Scientist at Zykrr Technologies Pvt Ltd (2024-05 – 2025-06)
- Built and launched a multi-agent chatbot using RAG, ReAct, and vector databases, reducing manual data-analysis time by 90% for enterprise clients.
- Designed and deployed a call-analytics system using a self-hosted Whisper model, cutting QA audit time by 83% (from three hours to 30 minutes per call).
- Led LLaMA 3.1 QLoRA fine-tuning to create specialized adapters that outperformed the base model by 30% on domain tasks and reduced token costs by 28%.
Senior Manager Data Scientist at Zykrr Technologies Pvt Ltd (2023-05 – 2024-05)
- Pioneered the company's GenAI strategy by building 3+ high-impact prototypes, securing executive buy-in and influencing the roadmap for its flagship chatbot product.
- Led predictive-analytics POCs that defined product direction and contributed to a 15% increase in client upsell opportunities.
Manager Data Scientist at Zykrr Technologies Pvt Ltd (2022-05 – 2023-05)
- Owned end-to-end development of the Key Driver Analysis product, used by 70%+ of enterprise clients and contributing $50K in new ARR.
- Engineered a T5-based NLP feature that extracts actionable insights from text, saving clients an average of 10 hours per week.
Senior Data Scientist at Zykrr Technologies Pvt Ltd (2021-05 – 2022-05)
- Reduced inference costs by 50% by replacing BERT with quantized DistilBERT models with negligible impact on accuracy.
- Fixed and merged a critical bug in the simpletransformers library, restoring NER functionality for the open-source community.
Data Scientist at Zykrr Technologies Pvt Ltd (2019-06 – 2021-05)
Internship: Apr 2019 - Jun 2019; promoted to full-time Data Scientist.
- Migrated the legacy analytics engine from TF-IDF to BERT, improving the core model's F1-score by 18% and strengthening robustness on unseen data through adversarial and behavioral testing.
- Built a Redis- and AWS-based data pipeline processing 10K+ text records per hour with zero data loss.
